Linux-mm Archive on lore.kernel.org
 help / color / Atom feed
From: Kirill Tkhai <ktkhai@virtuozzo.com>
To: shakeelb@google.com, akpm@linux-foundation.org,
	vdavydov.dev@gmail.com, linux-kernel@vger.kernel.org,
	linux-mm@kvack.org, aryabinin@virtuozzo.com,
	ktkhai@virtuozzo.com
Subject: [PATCH v2-REBASED] mm: Make count list_lru_one::nr_items lockless
Date: Thu, 30 Nov 2017 13:46:30 +0300
Message-ID: <151203869520.3915.2587549826865799173.stgit@localhost.localdomain> (raw)

During the reclaiming slab of a memcg, shrink_slab iterates
over all registered shrinkers in the system, and tries to count
and consume objects related to the cgroup. In case of memory
pressure, this behaves bad: I observe high system time and
time spent in list_lru_count_one() for many processes on RHEL7
kernel.

This patch makes list_lru_node::memcg_lrus rcu protected, that
allows to skip taking spinlock in list_lru_count_one().

Shakeel Butt with the patch observes signify perf graph change. He says:

========================================================================
Setup: running a fork-bomb in a memcg of 200MiB on a 8GiB and 4 vcpu
VM and recording the trace with 'perf record -g -a'.

The trace without the patch:

+  34.19%     fb.sh  [kernel.kallsyms]  [k] queued_spin_lock_slowpath
+  30.77%     fb.sh  [kernel.kallsyms]  [k] _raw_spin_lock
+   3.53%     fb.sh  [kernel.kallsyms]  [k] list_lru_count_one
+   2.26%     fb.sh  [kernel.kallsyms]  [k] super_cache_count
+   1.68%     fb.sh  [kernel.kallsyms]  [k] shrink_slab
+   0.59%     fb.sh  [kernel.kallsyms]  [k] down_read_trylock
+   0.48%     fb.sh  [kernel.kallsyms]  [k] _raw_spin_unlock_irqrestore
+   0.38%     fb.sh  [kernel.kallsyms]  [k] shrink_node_memcg
+   0.32%     fb.sh  [kernel.kallsyms]  [k] queue_work_on
+   0.26%     fb.sh  [kernel.kallsyms]  [k] count_shadow_nodes

With the patch:

+   0.16%     swapper  [kernel.kallsyms]    [k] default_idle
+   0.13%     oom_reaper  [kernel.kallsyms]    [k] mutex_spin_on_owner
+   0.05%     perf  [kernel.kallsyms]    [k] copy_user_generic_string
+   0.05%     init.real  [kernel.kallsyms]    [k] wait_consider_task
+   0.05%     kworker/0:0  [kernel.kallsyms]    [k] finish_task_switch
+   0.04%     kworker/2:1  [kernel.kallsyms]    [k] finish_task_switch
+   0.04%     kworker/3:1  [kernel.kallsyms]    [k] finish_task_switch
+   0.04%     kworker/1:0  [kernel.kallsyms]    [k] finish_task_switch
+   0.03%     binary  [kernel.kallsyms]    [k] copy_page
========================================================================

Thanks Shakeel for the testing.

Tested-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>

v2: Rebase on kvmalloc() and kvfree(). Name of kvfree_rcu()
was choosen to help not to skip this place, when someone
will implement such the global interface.
---
 include/linux/list_lru.h |    3 +-
 mm/list_lru.c            |   67 +++++++++++++++++++++++++++++++---------------
 2 files changed, 47 insertions(+), 23 deletions(-)

diff --git a/include/linux/list_lru.h b/include/linux/list_lru.h
index bb8129a3474d..96def9d15b1b 100644
--- a/include/linux/list_lru.h
+++ b/include/linux/list_lru.h
@@ -32,6 +32,7 @@ struct list_lru_one {
 };
 
 struct list_lru_memcg {
+	struct rcu_head		rcu;
 	/* array of per cgroup lists, indexed by memcg_cache_id */
 	struct list_lru_one	*lru[0];
 };
@@ -43,7 +44,7 @@ struct list_lru_node {
 	struct list_lru_one	lru;
 #if defined(CONFIG_MEMCG) && !defined(CONFIG_SLOB)
 	/* for cgroup aware lrus points to per cgroup lists, otherwise NULL */
-	struct list_lru_memcg	*memcg_lrus;
+	struct list_lru_memcg	__rcu *memcg_lrus;
 #endif
 	long nr_items;
 } ____cacheline_aligned_in_smp;
diff --git a/mm/list_lru.c b/mm/list_lru.c
index f141f0c80ff3..dc42a964b896 100644
--- a/mm/list_lru.c
+++ b/mm/list_lru.c
@@ -52,14 +52,15 @@ static inline bool list_lru_memcg_aware(struct list_lru *lru)
 static inline struct list_lru_one *
 list_lru_from_memcg_idx(struct list_lru_node *nlru, int idx)
 {
+	struct list_lru_memcg *memcg_lrus;
 	/*
-	 * The lock protects the array of per cgroup lists from relocation
-	 * (see memcg_update_list_lru_node).
+	 * Either lock or RCU protects the array of per cgroup lists
+	 * from relocation (see memcg_update_list_lru_node).
 	 */
-	lockdep_assert_held(&nlru->lock);
-	if (nlru->memcg_lrus && idx >= 0)
-		return nlru->memcg_lrus->lru[idx];
-
+	memcg_lrus = rcu_dereference_check(nlru->memcg_lrus,
+					   lockdep_is_held(&nlru->lock));
+	if (memcg_lrus && idx >= 0)
+		return memcg_lrus->lru[idx];
 	return &nlru->lru;
 }
 
@@ -168,10 +169,10 @@ static unsigned long __list_lru_count_one(struct list_lru *lru,
 	struct list_lru_one *l;
 	unsigned long count;
 
-	spin_lock(&nlru->lock);
+	rcu_read_lock();
 	l = list_lru_from_memcg_idx(nlru, memcg_idx);
 	count = l->nr_items;
-	spin_unlock(&nlru->lock);
+	rcu_read_unlock();
 
 	return count;
 }
@@ -323,24 +324,41 @@ static int __memcg_init_list_lru_node(struct list_lru_memcg *memcg_lrus,
 
 static int memcg_init_list_lru_node(struct list_lru_node *nlru)
 {
+	struct list_lru_memcg *memcg_lrus;
 	int size = memcg_nr_cache_ids;
 
-	nlru->memcg_lrus = kvmalloc(size * sizeof(void *), GFP_KERNEL);
-	if (!nlru->memcg_lrus)
+	memcg_lrus = kvmalloc(sizeof(*memcg_lrus) +
+			      size * sizeof(void *), GFP_KERNEL);
+	if (!memcg_lrus)
 		return -ENOMEM;
 
-	if (__memcg_init_list_lru_node(nlru->memcg_lrus, 0, size)) {
-		kvfree(nlru->memcg_lrus);
+	if (__memcg_init_list_lru_node(memcg_lrus, 0, size)) {
+		kvfree(memcg_lrus);
 		return -ENOMEM;
 	}
+	RCU_INIT_POINTER(nlru->memcg_lrus, memcg_lrus);
 
 	return 0;
 }
 
 static void memcg_destroy_list_lru_node(struct list_lru_node *nlru)
 {
-	__memcg_destroy_list_lru_node(nlru->memcg_lrus, 0, memcg_nr_cache_ids);
-	kvfree(nlru->memcg_lrus);
+	struct list_lru_memcg *memcg_lrus;
+	/*
+	 * This is called when shrinker has already been unregistered,
+	 * and nobody can use it. So, there is no need to use kvfree_rcu().
+	 */
+	memcg_lrus = rcu_dereference_protected(nlru->memcg_lrus, true);
+	__memcg_destroy_list_lru_node(memcg_lrus, 0, memcg_nr_cache_ids);
+	kvfree(memcg_lrus);
+}
+
+static void kvfree_rcu(struct rcu_head *head)
+{
+	struct list_lru_memcg *mlru;
+
+	mlru = container_of(head, struct list_lru_memcg, rcu);
+	kvfree(mlru);
 }
 
 static int memcg_update_list_lru_node(struct list_lru_node *nlru,
@@ -350,8 +368,9 @@ static int memcg_update_list_lru_node(struct list_lru_node *nlru,
 
 	BUG_ON(old_size > new_size);
 
-	old = nlru->memcg_lrus;
-	new = kvmalloc(new_size * sizeof(void *), GFP_KERNEL);
+	old = rcu_dereference_protected(nlru->memcg_lrus,
+					lockdep_is_held(&list_lrus_mutex));
+	new = kvmalloc(sizeof(*new) + new_size * sizeof(void *), GFP_KERNEL);
 	if (!new)
 		return -ENOMEM;
 
@@ -360,29 +379,33 @@ static int memcg_update_list_lru_node(struct list_lru_node *nlru,
 		return -ENOMEM;
 	}
 
-	memcpy(new, old, old_size * sizeof(void *));
+	memcpy(&new->lru, &old->lru, old_size * sizeof(void *));
 
 	/*
-	 * The lock guarantees that we won't race with a reader
-	 * (see list_lru_from_memcg_idx).
+	 * The locking below allows readers that hold nlru->lock avoid taking
+	 * rcu_read_lock (see list_lru_from_memcg_idx).
 	 *
 	 * Since list_lru_{add,del} may be called under an IRQ-safe lock,
 	 * we have to use IRQ-safe primitives here to avoid deadlock.
 	 */
 	spin_lock_irq(&nlru->lock);
-	nlru->memcg_lrus = new;
+	rcu_assign_pointer(nlru->memcg_lrus, new);
 	spin_unlock_irq(&nlru->lock);
 
-	kvfree(old);
+	call_rcu(&old->rcu, kvfree_rcu);
 	return 0;
 }
 
 static void memcg_cancel_update_list_lru_node(struct list_lru_node *nlru,
 					      int old_size, int new_size)
 {
+	struct list_lru_memcg *memcg_lrus;
+
+	memcg_lrus = rcu_dereference_protected(nlru->memcg_lrus,
+					       lockdep_is_held(&list_lrus_mutex));
 	/* do not bother shrinking the array back to the old size, because we
 	 * cannot handle allocation failures here */
-	__memcg_destroy_list_lru_node(nlru->memcg_lrus, old_size, new_size);
+	__memcg_destroy_list_lru_node(memcg_lrus, old_size, new_size);
 }
 
 static int memcg_init_list_lru(struct list_lru *lru, bool memcg_aware)

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

                 reply index

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=151203869520.3915.2587549826865799173.stgit@localhost.localdomain \
    --to=ktkhai@virtuozzo.com \
    --cc=akpm@linux-foundation.org \
    --cc=aryabinin@virtuozzo.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=shakeelb@google.com \
    --cc=vdavydov.dev@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-mm Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-mm/0 linux-mm/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-mm linux-mm/ https://lore.kernel.org/linux-mm \
		linux-mm@kvack.org
	public-inbox-index linux-mm

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kvack.linux-mm


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git